In this paper considering a new human gait recognition system based on Radon transform which gives a high precision recognition rate. Feature extraction in this work is based on the Radon transform of binary silhouettes .In this paper for each gait sequence the transformed silhouettes are used after background estimation and human detection to make each related templates. Set of all templates is used to subspace projection by PCA method. Consequently feature vectors are used to multi layer perceptron neural